yolov3 custom obj detection using process.py seperate train and test annotation txt and image files
how to download and use darknet tutorial below
• Video
convert .py download link look at the comments pin
to fix errors and install BBox Label Tool in linux mint and ubantu
( • fix errors and install BBox Label To... )
converting into files into yolo format the convert.py go to this video for explanation and code
( • yolo v3 custom object detection in bb... )
getting crystal clear explanation on yolo custum object detection go to this site below
(https://timebutt.github.io/static/how...)
to get the codes ,weights and cfg files go to this site below
(https://github.com/unsky/yolo-for-win...)
( https://github.com/AlexeyAB/darknet )
the darknet code to start training given below
/darknet detector train /home/saktheeswaran/darknet/build/darknet/x64/data/obj.data /home/saktheeswaran/darknet/build/darknet/x64/cfg/yolo-obj.cfg /home/saktheeswaran/darknet/build/darknet/x64/data/darknet19_448.conv.23
for easy and custom object detection files go to this site
can use this cfg file as easy alternaive to the above
https://github.com/Microsoft/VoTT/blo...
https://github.com/AlexeyAB/Yolo_mark...
How to Use Custom Tiny YOLOv2 Model on Android
http://seangtkelley.me/blog/2017/12/2...
yolo train on windows
https://jumabek.wordpress.com/2017/03...
now you can detect shapes also with yolo
Straight To Shapes: Real-Time Detection of Encoded Shapes
https://www.robots.ox.ac.uk/~tvg/proj...
for easy annotation for yolo and to make easy annotation and to convert it to yol format use these software below these saves you a lot time and yoy can skip the process of using converting.py if you have these software and very easy to use because they directly convert it to yolo format here are the website just download install and use them these will reall save you a lot time the websites are below
this first one is the best software the yolo boobs only thing youy have to do is you have to name the files as 1,2,3,4,5 and
save a files with the classes and what you are custom detecting as dog ,cat , and etc and it is important and madatory to save these file as classes.txt or else you cat annotate and and cant convert to yolo format here isthe code and the websiet this is easy as pie and just run the boobs.html and select the classes.txt now annotate and save it you have your yolo format for custom object detection without any pain
https://github.com/drainingsun/boobs
/ boobs-yolo-bbox-annotation-tool
or use this yolo -mark another tool which save times as boobs the yolo mark
https://github.com/AlexeyAB/Yolo_mark
* : ) so now you can cont on boobs for yolo : )*********
friends for a seem less gpu acceleration
try to make the file in vim as command given below
$ vi Makefile
now it gets into the vim editor now press insert in your keyboard and now change the GPU = 0 to GPU =1 and CUDNN=1 and OPENCV =1 then
get to the bottom of the vim editor and press esc
and type
:wq and press enter
now it gets out of the vim editor and now
type
$ make -j4
let it finish and now you can start the training and make sure you have entered the path of test and train files correctly then for cfg best to choose tiny-yolo.cfg or yolo-voc.2.0.cfg and change the classes as one or two according to the number of objects you want to detect and change the filters above the classes to filter=35 for recogonizing 2 and filter 18 for 1
and type the following command for training after training give the test commands the train and test commands are below
for trainig give the command with correct path below
$ ./darknet detector train /home/saktheeswaran/darknet/data/obj.data /home/saktheeswaran/darknet/cfg/tiny-yolo.cfg /home/saktheeswaran/darknet/darknet19_448.conv.23
for testing put the weights and start the training as below
./darknet detector test /home/saktheeswaran/darknet/data/obj.data /home/saktheeswaran/darknet/cfg/yolo-voc.2.0.cfg yolo-voc_100.weights data/5.JPEG
these commands depend on the user name and location of these files
Also wanted to ask, Can I use images that have several classes in them, in your case an image with a bird and a aeroplane? (marking different rectangles with different class tags. can this be in the same image?)
Yes, you can use image with marked many different object on the same image. Also you can use image without objects at all, I use it to avoid false-positive detections.
Is this normal. do I have a problem?
There is no problem. But for good detection you should use ~500 - 2000 images per class. And you should train 2000 iterations per class. So on CPU (without GPU) you will train 4000 iterations (for 2 classes) very much time :)
Смотрите видео yolov3 custom obj detection using process.py seperate train and test annotation txt and image files онлайн без регистрации, длительностью часов минут секунд в хорошем качестве. Это видео добавил пользователь RANDOM NEURAL MONK 29 Май 2018, не забудьте поделиться им ссылкой с друзьями и знакомыми, на нашем сайте его посмотрели 11,310 раз и оно понравилось 69 людям.